bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Profilin 1 Controls the Assembly, Organization, and Dynamics of Leading Edge Structures Through Internetwork Competition and Collaboration

Kristen Skruber1, Peyton V. Warp1, Rachael Shklyarov1, James D. Thomas2, Maurice S. Swanson2, Jessica L. Henty-Ridilla3, Tracy-Ann Read1, and Eric A. Vitriol1*

1Department of Anatomy and Biology, University of Florida, College of Medicine, Gainesville, FL, USA, 32610 2Department of Molecular Genetics and Microbiology, Center for NeuroGenetics and the Genetics Institute, University of Florida, College of Medicine, Gainesville, FL, USA, 32610 3Department of Cell and Developmental Biology, SUNY Upstate Medical University, NY, USA, 13210

*To whom correspondence should be addressed: Eric Vitriol, [email protected]

Abstract How actin monomers are distributed to different networks remains poorly understood. One emerging concept is that the monomer pool is limited and heterogenous, causing biased assembly and internetwork competition. However, most knowledge regarding monomer distribution comes from studies where competing networks are discrete. In metazoans, many actin-based structures are complex, containing competing networks that overlap and are functionally interdependent. Addressing how monomers control the assembly and organization of these complex structures is critical to understanding how actin functions in cells. Here, we identify the monomer-binding (PFN1) as a major determinant of actin assembly, organization, and network homeostasis in mammalian cells. At the leading edge, PFN1 controls the localization and activity of the assembly factors Arp2/3 and Mena/VASP, with discrete stages of internetwork competition and collaboration occurring at different PFN1 concentrations. This causes substantial changes to leading edge actin architecture and the types of structures that form there. ______

Introduction In cells, most actin monomers are bound to profilin (15). Profilin prevents spontaneous nucleation and directs To divide, move, and communicate, cells rely on a monomers to the fast-growing (barbed) ends of actin dynamic actin that can rapidly assemble and filaments (15, 16). Profilin also interacts with (17– adapt. This is achieved by the polymerization of actin 21) and Mena/VASP (7, 22, 23) at barbed ends and monomers into filaments, the construction of large enhances their ability to polymerize actin. Although filament networks, and the disassembly of these networks profilin has been shown to suppress branching of the back into monomers. To meet the demands of actin multi-component assembly factor Arp2/3 (6, 7, 24, 25), it network assembly, cells maintain a large monomer can supply monomers to Arp2/3-driven networks through reserve (1–4). However, several factors complicate how interactions with WASP-family (26–28). Thus monomers are distributed to different actin structures formins, Mena/VASP, and Arp2/3 all utilize profilin-actin within the cell (5). For example, monomers can undergo (19, 20, 22, 28–30), which can cause internetwork biased assembly into specific networks through competition (6, 7). The concept of monomer competition interactions with polymerases and monomer-binding has largely been inferred from experimental systems proteins (6, 7). Monomers can also be subcellularly where actin networks are spatially and functionally localized (8, 9) or recycled back into the structures which distinct. These include in vivo experiments in yeast, where they originated from (10). Actin isoforms can exhibit monomers polymerize into either Arp2/3-driven patches differential localization (11), dynamics (12), and or formin-based cables (31). Metazoan cells, however, regulation by post-translational modifications (13). Finally, contain actin structures where complex filament the monomer/filament ratio is in homeostasis (14), which architectures are constructed by multiple assembly causes networks competing for the same monomers to factors. Two examples of this are the lamellipodia, a alter their growth based on the activity or expression level protrusive structure at the leading edge of cells which is of different assembly factors (6, 7, 14). Thus, the rules made of dendritic and linear actin networks assembled by that govern how monomers are allocated throughout the Arp2/3, formins, and Mena/VASP (32–35) and filopodia, cell are complicated and many of the details still need to the finger-like actin projections that extend beyond the be uncovered.

Skruber et al. | bioRχiv | Nov. 19, 2019 | 1 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. lamellipodia (34). It is poorly understood how the competing for monomers overlap and are functionally monomer pool regulates the assembly and organization interdependent. of these types of actin superstructures, where networks

Figure 1. PFN1 controls global actin polymerization and monomer/filament homeostasis. (A) Western blot for profilin 1 (PFN1) of control (Ctrl) and PFN1 knockout (KO) CAD cells transfected with GFP or GFP- PFN1 (PFN1). (B) Representative images of the actin cytoskeleton in control and PFN1 KO cells transfected with GFP or GFP-PFN1. Actin filaments were labeled with Alexa568-phalloidin. Scale bar, 10 µm. (C) Western blot for actin in control and PFN1 KO cells expressing GFP or GFP-PFN1. Cell lysates were subjected to ultracentrifugation to separate monomers and filaments. Actin monomers (G-actin, designated as ‘G’) remain in the supernatant while filaments (F-actin, designated as ‘F’) sediment to the pellet. (Legend continues on next page).

Skruber et al. | bioRχiv | Nov. 19, 2019 | 2 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 1 (continued). PFN1 controls global actin polymerization and monomer/filament homeostasis. (D) Quantification of the G/F-actin ratio from (C). Individual data points are plotted along with the mean and 95% confidence intervals. Number of biological replicates is as follows: control + GFP (n = 6), PFN1 KO + GFP (n = 8), and PFN1 KO + GFP-PFN1 (n = 6). (E) Control and PFN1 KO cells transfected with GFP, GFP-PFN1 (GFP-PFN1WT), or the non-actin binding PFN1 mutant R88E (GFP-PFN1R88E). Actin filaments have been labeled with Alexa568- phalloidin. The images are scaled identically and pseudocolored based on the included lookup table to convey relative fluorescent intensities. Scale bar, 10 µm. (F) Quantification of mean Alexa568-phalloidin intensity in control and KO cells expressing GFP, GFP-PFN1R88E (R88E), or GFP-PFN1 (WT). Data are plotted relative to CTRL cells expressing GFP. For control cells expressing GFP, GFP-PFNR88E or GFP-PFN1WT, n = 475, 166, and 206, respectively. For PFN1 KO cells expressing GFP, GFP-PFNR88E or GFP-PFN1WT, n = 446, 206, and 443, respectively. (G) Correlation between GFP and phalloidin intensity for cells in (E). Intensities were normalized to the mean of each data set. (H-J) Representative images and quantification of mean Alexa568- phalloidin intensity in control and PFN1 KO cells, where Arp2/3, formins, or Mena/VASP were inhibited. The images are scaled identically and pseudocolored based on the included lookup table to convey relative fluorescent intensities. Scale bar, 10 µm. (H) Control and PFN1 KO cells were pre- treated with control (CK-689) or Arp2/3 (CK-666) small molecule inhibitors for 1 hour prior to fixation and labeling with Alexa568-phalloidin. For control cells n = 154 for CK-689 and 93 for CK-666. For PFN1 KO cells n = 111 for CK-689 and 119 for CK-666. (I) Control and PFN1 KO cells were pre-treated with vehicle control (DMSO) or the pan-formin small molecule inhibitor SMIFH2 for 20 min prior to fixation and labeling with Alexa568-phalloidin. For control cells n = 140 for DMSO and n = 130 for SMIFH2. For PFN1 KO cells n = 160 for DMSO and n = 159 for SMIFH2. (J) Control and PFN1 KO cells were transfected with the mitochondria sequestering Mena/VASP construct FP4 mito, or its AP4 mito control for 16 hour prior to fixation and labeling with Alexa568-phalloidin. For control cells n = 89 for AP4 mito and n = 94 for FP4 mito. For PFN1 KO cells n = 79 for AP4 mito and n = 96 for FP4 mito. Box-and-whisker plots in D,F,H,I,J denote 95th (top whisker), 75th (top edge of box), 25th (bottom edge of box), and 10th (bottom whisker) percentiles and the median (bold line in box). p values plotted relative to control + GFP, unless otherwise indicated. **** indicates p ≤ 0.0001, *** p ≤ 0.001, n.s. = not significant (p > 0.05). p values were generated by ANOVA followed by Tukey’s post hoc test (comparison of ³ 3 conditions).

In this study, we dissect how actin assembly factors and filament-containing cellular fractions revealed that the collectively construct complex actin networks through ratio of monomeric to polymerized actin is 3:1 in PFN1 KO Profilin 1 (PFN1). Using a PFN1 knockout/rescue cells (Fig. 1C, D), and can be rescued to the control cell experimental paradigm, we demonstrate that PFN1 ratio by expressing GFP-PFN1 (Fig. 1C, D). Quantitative controls the majority of actin polymerization, determines image analysis of cells labeled with fluorescent phalloidin the monomer to filament ratio set point, and facilitates was then used to measure relative amounts of actin homeostatic interplay between different actin networks. filaments. Filament levels in PFN1 KO cells were reduced These properties extend to PFN1’s regulation of actin at more than 50% in comparison to control cells (Fig. 1E, F). the leading edge, where it coordinates the localization and This reduction was rescued by expression of wild-type activity of different actin assembly factors in a GFP-PFN1 but not the non-actin binding R88E mutant concentration-dependent manner and determines which (38) (Fig. 1E, F). While knocking out PFN1 caused a slight leading edge structures assemble. Thus, monomer reduction in actin expression (Supplementary Fig. 1A), it distribution through PFN1 is a major determinant of actin was insufficient to explain the substantially larger assembly, organization, and homeostasis in cells and a decrease in polymerized actin levels in PFN1 KO cells critical regulator of complex actin architectures. (Fig. 1E, F) and the reciprocal increase in the monomer/filament ratio (Fig. 1C, D). Additionally, Results and Discussion supplying PFN1 KO cells with more actin by expressing EGFP-β-actin did not increase actin polymerization PFN1 controls global actin polymerization and (Supplementary Fig. 1B, C), verifying that filament levels monomer/filament homeostasis are low because the monomers cannot polymerize in the We knocked out the PFN1 gene with CRISPR/Cas9 in absence of PFN1. The substantial inability of actin to Cath.a differentiated (CAD) cells. CAD cells were chosen polymerize in PFN1 KO cells is likely due to a combination for this study because we have previously characterized of an increase in filament capping (39), inhibition of or how their monomer pool influences actin network decrease in barbed-end polymerase activity (20, 22), a behavior (8, 10, 12, 36) and because PFN1 is the decrease in the nucleotide exchange rate of actin (40), predominant profilin isoform. PFN2 is expressed and an increase in sequestering by Thymosin β4 (41). approximately 10-fold less than PFN1, while PFN3 and PFN4 are not expressed. Importantly, PFN2 expression Overexpressing PFN1 in control cells caused an increase doesn’t change upon deletion of PFN1 (Supplementary in actin filaments (Fig. 1E, F). We postulated that PFN1 Table 1), which is consistent with studies showing that concentration was the predominant factor which these isoforms cannot compensate for each other (37). determined the monomer to filament ratio and that global Additionally, very few other actin-binding proteins are actin polymerization would scale with PFN1 expression differentially expressed when PFN1 is knocked out levels. To test this, we performed a correlation analysis (Supplementary Table 2). Thus, this is a strong cell line to between PFN1 expression and polymerized actin. In study how PFN1 regulates the actin cytoskeleton. Loss of control and KO cells, there was a significant, positive, PFN1 was verified by western blot (Fig. 1A, B). The actin linear correlation between actin filament levels and PFN1 cytoskeleton is drastically altered in PFN1 KO cells, which expression, but not with expression of GFP or GFP- can be rescued with physiologically relevant expression PFN1R88E (Fig. 1G). Interestingly, the R88E mutant had a levels of GFP-PFN1 (Fig. 1A, B). slightly dominant negative effect on actin polymerization, most likely due to binding poly-L- residues but not We first sought to determine how actin polymerization and actin (38). These results demonstrate that PFN1 is not monomer/filament homeostasis are disrupted in the only needed for actin to polymerize (Fig. 1C-F), but that absence of PFN1. Western blot quantification of monomer Skruber et al. | bioRχiv | Nov. 19, 2019 | 3 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. its expression level determines the monomer/filament internetwork homeostasis, we measured polymerized homeostasis set point. actin levels after specific actin assembly factors were inhibited. As predicted by previous work (6, 7), global Monomer/filament homeostasis has been shown in yeast, filament levels were maintained in control cells after where inhibiting Arp2/3-driven actin patches causes a inhibition of Arp2/3 or formins (Fig. 1H, I). However, reciprocal increase of formin-based actin cables (6, 14). inhibiting Arp2/3 or formins in PFN1 KO cells caused a To determine if PFN1 was needed to maintain significant reduction in actin filaments (Fig. 1H, I),

Skruber et al. | bioRχiv | Nov. 19, 2019 | 4 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Figure 2 (previous page). PFN1 controls assembly and organization of actin at the leading edge. (A) Representative confocal super-resolution images of control and PFN1 KO cells expressing GFP, GFP-PFN1R88E, or GFP-PFN1WT and labeled with Alexa568-phalloidin. Insets highlight actin at the leading edge. Scale bar, 10 µm. (B) Ratio between the sum intensity of the leading edge and the cellular interior of control and PFN1 KO cells expressing GFP, GFP-PFN1R88E (R88E), or GFP-PFN1WT (WT) and labeled with Alexa568-phalloidin. For all conditions n = 25. (C) Linescan analysis of the leading edge of control and PFN1 KO cells expressing GFP, GFP-PFN1R88E, or GFP-PFN1WT labeled with Alexa568-phalloidin. The transparent bands depict 95% confidence intervals. For all conditions, n = 400 lines drawn from 20 cells. (D) Representative kymographs of lamellipodia retrograde flow in control and PFN1 KO cells expressing Lifeact-mRuby and either GFP, GFP-PFN1R88E (R88E), or GFP- PFN1WT (WT). (E) Quantification of retrograde flow from kymographs as depicted in (D). For control cells n = 170 measurements from 17 cells. For PFN1 KO cells n = 310 measurements from 31 cells for GFP; n = 100 measurements from 10 cells for R88E; and n = 180 measurements from 18 cells for WT. (F) Confocal images showing actin filaments (Phalloidin) and barbed ends (Rhodamine Actin) in control and PFN1 KO cells. Cells were gently permeabilized and incubated with rhodamine actin in polymerization buffer for 60 s to label actin filament barbed ends. They were then fixed and incubated with Alexa488-phalloidin to label actin filaments. Black scale bar, 10 µm. Red scale bar in inset, 1 µm. (G) Quantification of barbed ends in control and PFN1 KO cells by measuring sum rhodamine actin fluorescence. Fluorescent intensity is plotted relative to control cells. For control cells n = 30, for PFN1 KO cells, n = 25. (H–J) Quantification of linear arrays at the leading edge of control and PFN1 KO cells expressing GFP, GFP-PFN1R88E or GFP-PFN1WT. (H) Representative confocal super-resolution images of the leading edge of control and PFN1 KO cells expressing GFP, GFP-PFN1R88E, or GFP-PFN1WT and labeled with Alexa568-phalloidin. Results from linear array segmentation analysis are outlined in red. Scale bar, 5 µm. (I and J) Linear array area and density measurements of control cells expressing GFP (n = 5075 arrays, 28 cells), GFP-PFN1R88E (n = 1937 arrays, 17 cells) or GFP-PFN1WT (n = 2813 arrays, 20 cells) and PFN1 KO cells expressing GFP (n = 1559 arrays, 15 cells), GFP-PFN1R88E (n = 1903 arrays, 25 cells) or GFP-PFN1WT (n = 2688 arrays, 15 cells). Area measures the size of individual arrays. Density was measured by normalizing the linear array count to the cell perimeter.

Box-and-whisker plots in B,E,G,I,J denote 95th (top whisker), 75th (top edge of box), 25th (bottom edge of box), and 10th (bottom whisker) percentiles and the median (bold line in box). p values plotted relative to control + GFP, unless otherwise indicated. **** indicates p ≤ 0.0001, *** p ≤ 0.001, n.s. = not significant (p > 0.05). p values were generated by either a two-tailed student’s t-test (comparison of two conditions) or by ANOVA followed by Tukey’s post hoc test (comparison of ³ 3 conditions). For (I) Dunn’s post hoc test was used to generate p-values after normality of the data was assessed. revealing an inability to undergo compensatory network expressing GFP-PFN1WT but not GFP-PFN1R88E (Fig. 2A- assembly. Furthermore, inducing loss of function of E). One factor that drives retrograde flow rate is the Mena/VASP proteins by targeting them to mitochondria number of actively polymerizing filaments (42). To confirm (Supplementary Fig. 2) did not alter actin polymerization that actin polymerization was reduced at the leading edge levels in PFN1 KO cells (Fig. 1J), but did reduce actin of PFN1 KO lamellipodia, we labeled sites of active filaments in control cells. This result supports prior work polymerization (43) (see Material and Methods for details) demonstrating that Mena/VASP requires profilin-actin for and found that PFN1 KO cells have approximately 50% its polymerase activity (22, 29). Because of the potential fewer actively polymerizing filaments than control cells for compensatory mechanisms to occur in the time (Fig. 2F, G). Again, this is likely due to a decrease in needed to express the FP4-mito construct, these barbed end polymerase activity and an increase in experiments cannot be directly compared to the filament capping (39, 44). immediate inactivation of Arp2/3 and formin by small molecule inhibitors. However, these experiments do still To understand how PFN1 expression alters lamellipodia reveal that PFN1 plays an integral role in the assembly of architecture, we quantified linear filament arrays (see Arp2/3, formin, and Mena/VASP-based networks. Arp2/3 Materials and Methods for details), which are the and formins can function independently of PFN1, but their predominant products of formins and Mena/VASP at the internetwork monomer/filament homeostasis is not leading edge (33, 34, 45). In PFN1 KO cells, linear array maintained unless PFN1 is present. Actin assembly by area and density were both substantially reduced Mena/VASP is PFN1-dependent and not in homeostasis compared to control cells (Fig. 2H-J). The loss of linear with Arp2/3 and formin-based networks, demonstrating arrays in PFN1 KO cells is rescuable by expressing GFP- an exclusive use of PFN1-actin that does not transfer to PFN1WT but not GFP-PFN1R88E (Fig. 2H-J). PFN1 other the assembly factors. overexpression increased linear array area in control cells (Fig. 2H-J), demonstrating that the size of lamellipodia PFN1 controls assembly and organization of actin linear array networks is modulated by PFN1 at the leading edge concentration, in a manner similar to global actin polymerization (Fig. 1G). Interestingly, PFN1 over- Since PFN1 was shown to coordinate actin expression caused the arrays to grow larger (Fig. 2I) but polymerization by Arp2/3, formins, and Mena/VASP, we did not increase their density (Fig. 2J), suggesting that wanted to determine how it controlled the assembly and once a certain number of bundles has been assembled, organization of actin structures that depend on all three these actin networks can only expand by increasing in assembly factors: the lamellipodia (32–34) and filopodia size, perhaps due to space limitations and/or competition (convergent elongation ref). PFN1 KO cells have a for resources from Arp2/3-based dendritic networks (14, dramatically altered lamellipodia (Fig. 2A) with 46). approximately three-fold less actin than control cells (Fig. 2B). Linescan analysis revealed that PFN1 KO PFN1 coordinates Arp2/3 and Mena/VASP lamellipodia are also smaller (Fig. 2C). Live cell imaging localization and activity at the leading edge of cells expressing Lifeact-mRuby demonstrated that actin retrograde flow rates were reduced in PFN1 KO cells Next, we sought to understand how PFN1 controls Arp2/3 by approximately 50% (Fig. 2D, E). All lamellipodia and Mena/VASP activity in the lamellipodia. Immunocyto- phenotypes in PFN1 KO cells could be rescued by chemistry confirmed that Arp2/3 and Mena both localized

Skruber et al. | bioRχiv | Nov. 19, 2019 | 5 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. to the leading edge in control cells as expected (Fig. 3A- 3A-C). We confirmed that this difference in leading edge C). However, Arp2/3 is depleted from the leading edge of localization cannot be explained by differential expression PFN1 KO cells and Mena localization is increased (Fig.

Skruber et al. | bioRχiv | Nov. 19, 2019 | 6 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Figure 3 (previous page). PFN1 coordinates Arp2/3 and Mena/VASP localization and activity at the leading edge. (A) Representative images of ARPC2 and Mena immunolabeling in control and PFN1 KO cells. Scale bar, 10 µm. (B) Linescan analysis of ARPC2 immunolabeling at the leading edge of control and PFN1 KO cells expressing GFP or GFP-PFN1. The transparent bands depict 95% confidence intervals. For all conditions n = 300 linescans from 15 cells. (C) Linescan analysis of Mena immunolabeling at the leading edge of control and PFN1 KO cells expressing GFP or GFP-PFN1. The transparent bands depict 95% confidence intervals. For all conditions n = 400 linescans from 20 cells. (D) Representative confocal super-resolution images of the leading edge of control and PFN1 KO cells expressing AP4- or FP4-Mito and labeled with Alexa568-phalloidin. Scale bar, 5 µm. (E) Linescan analysis of the leading edge of control and PFN1 KO cells expressing AP4- or FP4-Mito, and then labeled with Alexa568-phalloidin. For control cells expressing AP4-Mito, 320 lines were drawn from 16 cells; for control cells expressing FP4-Mito 400 lines were drawn from 20 cells; for PFN1 KO cells expressing AP4-Mito and FP4-Mito, 300 lines were drawn from 15 cells. The transparent bands depict 95% confidence intervals. (F) Representative confocal super-resolution images of the leading edge of control and PFN1 KO cells expressing the MENA/VASP mitochondria-sequestering FP4-mito or its control AP4-mito. Cells expressed the DNA vectors for 16 hrs before they were fixed and labeled with Alexa568-phalloidin. Results from linear array segmentation analysis are outlined in red. Scale bar, 10 µm. (G and H) Linear array area and density measurements of control and PFN1 KO cells expressing AP4-mito or FP4-mito as depicted in (F). For control + AP4-mito, n =1928 arrays, 15 cells; For control + FP4-mito, n = 962 arrays, 20 cells; For PFN1 + AP4-mito, n = 853 arrays, 15 cells; For PFN1 + AP4-mito, n = 961 arrays, 15 cells. (I) Representative confocal super-resolution images of the leading edge of control and PFN1 KO cells that were pre-treated with control (CK-689) or Arp2/3 (CK-666) small molecule inhibitors for 1 hr prior to fixation and labeling with Alexa568-phalloidin. Results from linear array segmentation analysis are outlined in red. Scale bar 10 µm. (J and K) Linear array area and density measurements of control and PFN1 KO cells treated with CK-689 or CK-686 as depicted in (I). For control + CK-689, n = 3846 arrays, 25 cells; For control + CK-666, n = 4777 arrays, 30 cells; for PFN1 KO + CK-689, n = 2078 arrays, 29 cells; for PFN1 KO + CK-666, n = 2856 arrays, 23 cells. Area measures the size of individual linear arrays. Density was measured by normalizing the linear array count to the cell perimeter. (L) Representative kymographs of lamellipodia retrograde flow in control and PFN1 KO cells expressing Lifeact- mRuby and treated with CK-689 or CK-666 for 60 minutes prior to imaging. (M) Quantification of retrograde flow from kymographs as depicted in (L). For control + CK-689 and control + CK-666 n = 70 measurements, 7 cells; for PFN1 KO + CK-689 n = 80 measurements, 8 cells; for PFN1 KO + CK-666 n = 120 measurements, 12 cells. (N) Representative kymographs of lamellipodia retrograde flow in control cells expressing Lifeact-mRuby and either AP4- mito or FP4-mito. (O) Quantification of retrograde flow from kymographs as depicted in (N). For control + AP4-mito n =100 measurements, 10 cells; for control + FP4-mito n = 90 measurements, 9 cells. Box-and-whisker plots in E,H, M,O denote 95th (top whisker), 75th (top edge of box), 25th (bottom edge of box), and 10th (bottom whisker) percentiles and the median (bold line in box). Data in F,I are plotted as median with interquartile range. p values plotted relative to control, unless otherwise indicated. **** indicates p ≤ 0.0001, *** p ≤ 0.001, n.s. = not significant (p > 0.05). p values were generated by either a two-tailed student’s t-test (comparison of two conditions) or by ANOVA followed by Tukey’s post hoc test (comparison of ³ 3 conditions). For (G,J) Dunn’s post hoc test was used to generate p-values after normality of the data was assessed. of Arp2/3 or Mena (Supplementary Fig. 3). This result was density of linear arrays (Fig. 3F-H). However, there was surprising as it has previously been shown that reducing no change in linear array properties when FP4-mito was PFN1 expression enhanced Arp2/3 localization to the expressed in PFN1 KO cells (Fig. 3F-H). Together, these leading edge (7), supporting a role for PFN1 as an data indicate that leading edge-localized Mena is inert in inhibitor of Arp2/3 network assembly. However, in light of PFN1-depleted cells: it is not contributing to or inhibiting recent work showing that profilin transfers monomers to actin polymerization. These results also corroborate Arp2/3 networks via membrane-concentrated WASP- previous work showing that Mena requires profilin-actin family proteins (28), we may be seeing a similar role for for its polymerase activity (22, 29). PFN1 in our experimental system that could only be revealed with a complete loss-of-function. Leading edge actin architecture and Arp2/3 activity are defined by PFN1 concentration We have also shown that inducing loss of function by targeting Mena/VASP to mitochondria with FP4-mito To determine if PFN1-mediated internetwork competition (Supplementary Fig. 2) did not change total levels of F- between Arp2/3, formins, and Mena/VASP drives actin in PFN1 KO cells (Fig. 1G), suggesting that PFN1- complex actin network formation at the leading edge, we actin is required for Mena/VASP polymerase activity. used electroporation to introduce defined amounts of Thus, it was also perplexing that Mena concentrated at PFN1 protein into cells. Protein electroporation can be 6 the leading edge in PFN1’s absence. Mena’s localization performed on 10 cells simultaneously with a >99% there could indicate Mena is tethered to the plasma transfection rate. Additionally, by circumventing the membrane or localized to filament barbed ends (47). In normal biosynthetic pathway, proteins can be studied the latter case, if Mena cannot function as a polymerase under conditions that minimize a transcriptional response without profilin-actin (as indicated in Fig. 1G), it could be (49, 50). Using fluorescently labeled dextran, we acting as a capping protein by preventing barbed end demonstrated that the amount of delivered protein was growth. To address these possibilities, we used linescan linearly proportional to the bath concentration of protein in analysis of fluorescently labeled actin filaments to the electroporation chamber. Moreover, the variability in compare the intensity profiles in lamellipodia of control electroporation efficiency was low, allowing for discrete and PFN1 KO cells expressing FP4-mito. FP4-mito concentrations of delivered material (Supplementary Fig. expression in control cells severely depleted lamellipodial 4A-C). PFN1 protein could also be introduced into cells at actin in a manner that was nearly identical to knocking out defined concentrations, including physiological levels PFN1, while it had no additional effect on the lamellipodia (Supplementary Fig. 4D,E). In the following section and in of PFN1 KO cells (Fig. 3D, E). Additionally, we measured Figure 4, when we mention PFN1 concentration, we are linear arrays in the lamellipodia of FP4-mito expressing referring to the bath concentration of the electroporation cells. As predicted, inhibiting Mena/VASP at the leading chamber and not the amount protein delivered into cells. edge in control cells had significant effects on the size and Modulating the PFN1 concentration in PFN1 KO cells had dramatic effects on lamellipodia architecture. At 20 µM

Skruber et al. | bioRχiv | Nov. 19, 2019 | 7 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

PFN1, the lamellipodia was virtually eliminated and concentration, which completely rescued PFN1 protein caused the cells to send out numerous filopodia expression to control levels (Supplementary Fig. 4E), protrusions (Fig. 4A, B). At intermediate concentrations restored the size and architecture of the lamellipodia to (50 µM) of PFN1, the lamellipodia returned and filopodia strongly resembled control cells (Fig. 4A, B). Thus, the protrusions subsided (Fig. 4A, B). The 100 µM type and amount of actin filament structures that form at

Skruber et al. | bioRχiv | Nov. 19, 2019 | 8 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Figure 4 (previous page). Lamellipodia architecture and Arp2/3 activity are defined by the cellular PFN1 concentration. (A) Representative confocal super-resolution images of the leading edge of control and PFN1 KO cells after electroporation with the designated concentration of purified PFN1 and then labeled with Alexa568-phalloidin. Insets highlight actin architecture at the leading edge. Scale bar, 5 µm. (B) Linescan analysis of Alexa568-phalloidin fluorescence intensity at the leading edge of cells depicted in (A). Only regions containing a lamellipodia were analyzed. The transparent bands depict 95% confidence intervals. For all conditions, n = 400 linescans from 20 cells. (C and D) Linear array area and density measurements of control and PFN1 KO cells electroporated with the designated concentration of purified profilin 1 as depicted in (A). For control cells n = 5652 arrays, 27 cells. For PFN1 KO cells + 0 µM n = 2609 arrays, 27 cells. For PFN1 KO cells + 20 µM n = 3732 arrays, 34 cells. For PFN1 KO cells + 50 µM n = 5265 arrays, 35 cells. For PFN1 KO cells + 100 µM n = 4943 arrays, 30 cells. (E) Quantification of filopodia at the leading edge in control and PFN1 KO cells after electroporation with the designated concentration of purified PFN1 as depicted in (A). Filopodia were quantified by taking a ratio of the cell perimeter to the cell area. Higher ratios reflect more filopodia protrusions. For control cells n = 35. For PFN1 KO cells electroporated with 0 µM, 20 µM, 50 µM, and 100 µM n = 29, 36, 25, and 34, respectively. (F) Representative confocal images of control and PFN1 KO cells electroporated with the indicated concentration of purified profilin 1 and immunolabeled for Arp2/3 with an ARPC2 antibody. Scale bar, 10 µm. (G) Linescan analysis of ARPC2 fluorescence intensity at the leading edge of cells depicted in (F). The transparent bands depict 95% confidence intervals. For all conditions, n = 400 linescans from 20 cells. (H) Representative confocal super-resolution images of the leading edge of PFN1 KO cells after they were electroporated with the designated concentration of purified profilin 1, treated with CK-689 or CK-666 for 60 minutes, and then labeled with Alexa568-phalloidin. Scale bar, 5 µm. (I) Linear array area measurements of PFN1 KO cells electroporated with 20 or 100 µM of PFN1 protein and then treated with CK-689 or CK-686 as depicted in (H). For PFN1 KO + 20 µM PFN1, CK-689, n =1329 arrays, 15 cells; For PFN1 KO + 20 µM PFN1 + CK-666, n = 1511 arrays, 10 cells; for PFN1 KO + 100 µM profilin 1, CK-689 n = 1790 arrays, 10 cells; for PFN1 KO + 100 µM PFN1, CK-666, n = 1424 arrays, 10 cells. Area measures the size of each individual linear array. (J) Quantification of filopodia at the leading edge in PFN1 KO cells after electroporation with 20 or 100 µM of profilin 1 protein and then treatment with CK-689 or CK-686 as depicted in (H). Filopodia were quantified by taking a ratio of the cell perimeter to the cell area. Higher ratios reflect more filopodia protrusions. For PFN1 KO cells electroporated with 20 µM n=10, 15 for cells treated with CK-689 and CK-66, respectively. For PFN1 KO cells electroporated with 100 µM n = 10, 15 for cells treated with CK-689 and CK-666, respectively. (K) Linescan analysis of phalloidin fluorescence intensity at the leading edge of cells depicted in (H). The transparent bands depict 95% confidence intervals. For all conditions, n = 200 linescans from 10 cells. Lines were drawn in between arrays. Box-and-whisker plots in C,D,E,J,K denote 95th (top whisker), 75th (top edge of box), 25th (bottom edge of box), and 10th (bottom whisker) percentiles and the median (bold line in box). p values plotted relative to control + 0 µM profilin 1 protein, unless otherwise indicated. **** indicates p ≤ 0.0001, *** p ≤ 0.001, n.s. = not significant (p > 0.05). p values were generated by ANOVA followed by Tukey’s post hoc test (comparison of ³ 3 conditions). For (C,I) Dunn’s post hoc test was used to generate p-values after normality of the data was assessed. the leading edge are largely dependent on the availability on Arp2/3 activity, as suggested by the convergent of monomers bound to PFN1. elongation model (51). This model proposes an interdependence of filopodia generation on Arp2/3- Linear array analysis of the leading edge of PFN1 KO dependent nucleation, where Arp2/3 generates the base cells rescued with varying amounts of PFN1 further of the filament (51, 52). To test for convergent elongation, revealed how PFN1 controls leading edge actin we combined controlled PFN1 delivery with CK-666- architecture. Interestingly, the size and density of linear mediated inhibition of Arp2/3 (Fig. 4H-K). When PFN1 arrays increases with PFN1 concentration (Fig. 4E-G). concentration was limited to cause a strong induction of However, at lower PFN1 concentrations, the linear arrays filopodia, there was no difference in leading edge largely exist as filipodia-like protrusions. At higher PFN1 architecture after Arp2/3 inhibition, demonstrating these concentrations, these protrusions subside and the linear filipodia are Arp2/3-independent. Through Arp2/3 arrays exist primarily as filament bundles within the inhibition, we also show that the filopodia seen at low lamellipodia network (Fig. 4E). Measuring the cell’s PFN1 concentrations (Fig. 4A,H) are not the result of the perimeter/area ratio, which is significantly increased when dissolution of a dendritic network and subsequent filopodia are present, confirmed the biphasic regulation of unmasking of stable filopodia (53), but rather due to the filopodia by PFN1 concentration (Fig. 4H). Along with selective polymerization of specific structures. At 20 M previous experiments (Fig. 4A-D), this highlights the PFN1, CK-666 had no effect on linear filament arrays at sensitivity of leading edge actin architecture to the the leading edge (Fig. 4 I,J). Arp2/3 inhibition similarly had availability of PFN1-actin. Since Arp2/3 localization at the no effect on lamellipodia actin as measured by linescan leading edge is severely reduced in PFN1 KO cells (Fig. analysis of fluorescently-labeled actin filaments (Fig. 3A, B) and intermediate concentrations of PFN1 favor 4H,I). However, when cells were given 100 µM PFN1 and formation of filopodia over lamellipodia (Fig. 4A, B), we treated with CK-666, the actin structures that form are wanted to determine how PFN1 concentration affects highly affected by Arp2/3 inactivation: the lamellipodia is Arp2/3 localization to the leading edge. Surprisingly, we severely decreased (Fig. 4K) and large filopodia form found that the majority of Arp2/3 returns to the leading (Fig. 4I, J), demonstrating a shift toward assembly edge upon introduction of a low (20 µM) PFN1 through formins and Mena/VASP. These data corroborate concentration (Fig. 4C, D), despite no change in other studies showing PFN1 is crucial for homeostasis lamellipodia size in comparison to PFN1 KO cells (Fig. upon Arp2/3 inhibition (6, 7) but demonstrate an additional 4B). While Arp2/3 localization continued to increase with level of regulation where Arp2/3 activity also relies on higher PFN1 concentrations (Fig. 4C, D), 20 µM was PFN1. sufficient to recall the majority of Arp2/3 back to the leading edge. PFN1 KO cells are affected by Arp2/3 inactivation (Fig. 1H) indicating that Arp2/3 globally contributes to building Discovering a PFN1 concentration that potently actin networks without PFN1. Although at the leading stimulated the production of filopodia-like protrusions edge, Arp2/3 localization (Fig. 4F,G) and activity (Fig. allowed us to test whether their formation was dependent 3L,M) require PFN1. Polymerization of linear filament

Skruber et al. | bioRχiv | Nov. 19, 2019 | 9 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. arrays by PFN1, formins, and Mena/VASP (Fig. 3F-K) RNA was used as input to generate strand-specific, may be necessary to create Arp2/3 binding sites at the rRNA-depleted RNA-seq libraries using the KAPA leading edge, defining a collaboration between the stranded RNA-seq Kit with RiboErase HMR (Kapa different networks. This is reminiscent of filipodia/veil Biosystems). All steps were performed according to the motility first identified in neuronal growth cones, where manufacturer’s protocol except for the use of custom filopodia provide the initial step of membrane protrusions, Illumina-compatible index primers to allow multiplexing. followed by Arp2/3-based dendritic networks (54). Paired-end, 36 bp sequencing of the final libraries was However, low concentrations of PFN1 favor linear array performed using an Illumina NextSeq500. Gene network assembly (Fig. 4A-E), and prevent Arp2/3 from expression analysis was performed as previously assembling dendritic networks (Fig. 4H-K). Dendritic described (56). Briefly, reads were de-multiplex based on networks may rely less heavily on PFN1 than linear sample specific barcodes and mapped to the human networks do and thus may not be able to compete for genome (hg19) using OLego (57). Uniquely mapped profilin-actin. Higher PFN1 concentrations reduce this reads were assigned to genomic features and counted competition and allow all networks to form, though the using Quantas (58). TMM normalization and identification homeostatic setpoint can be modulated by increasing of differentially expressed genes ws computed using PFN1 (Fig. 3I-K, 4H-K, 100 µm). This result helps edgeR (59). Final gene lists were filtered (|log2 fold reconcile studies where PFN1 has been shown to both change| ≥ 1; adjusted P ≤ 0.01) to identify significant inhibit (7) and enhance (28) Arp2/3-based network changes. assembly. Thus, by carefully controlling protein concentration, we were able to determine how PFN1 Cell culture distributes monomers to different networks, even within Cath.-a-differentiated (CAD) cells (purchased from complex actin structures. We speculate that PFN1 Sigma-Aldrich) were cultured in DMEM/F12 medium expression levels will largely determine the types of actin (Gibco) supplemented with 8% fetal calf serum, 1% L- assembly that occur in a cell. Further, we have Glutamine, and 1% penicillin-streptomycin. Prior to demonstrated that the types of actin structures that form imaging, CAD cells were plated on coverslips coated with within a cell can be dictated solely by PFN1 availability. 10 μg/mL Laminin (Sigma-Aldrich). DMEM/F12 medium Thus, a cell could dramatically change its actin networks without phenol red (Gibco) supplemented with 15mM by modulating PFN1 gene expression or by using HEPES was used for live-cell imaging. CAD cells are a phosphorylation to reduce its ability to bind actin (55). unique mouse neuroblastoma cell line that differentiate into a neuronal-like cell morphology upon serum Methods withdrawal (60). We routinely use serum withdrawal to validate CAD cells by ensuring that they are able to DNA constructs undergo neuronal differentiation as evidenced by the The following DNA constructs were used in this study: formation of long (> 100 μm), narrow projections after 2 EGFP-PFN1 (Plasmid #56438, Addgene), Lifeact mRuby days. Cell lines were also routinely tested for mycoplasma (pN1-Lifeact-mRuby, provided by Roland Wedlich- using the Universal Detection Kit (ATCC). Soldner, Max-Planck Institute of Biochemistry), pEGFP- C1 EGFP β-actin, pMSCV EGFP-FP4 mito and EGFP- PFN1 KO cells were generated with CRISPR/Cas9 by AP4 mito (provided by Alpha Yap) (33, 47). EGFP- transfecting CAD cells with the constructs described PFN1R88E was generated from EGFP-PFN1 with site- above. One week after transfection, cells that were directed mutagenesis (Q5 New England Biolabs) using modified by CRISPR/Cas9 were selected with 10 μg/mL the primers AATGGATCTTGAAACCAAGAGCACC puromycin (Santa Cruz Biotechnology). This (forward) and GTAAATTCCCCG-TCTTGC (reverse). concentration was chosen as it kills 100% of cells that do Mutagenesis was confirmed by sequencing (Genewiz). not have the puromycin resistance gene within 24 hours EGFP-ß-actin was generated in a previous study (10). (10, 12, 36). Puromycin was removed 24 hours prior to PFN1 KO cells were generated with the pCRISPR-CG02 experiments that required transfection. vector (Genecopoeia) containing an sgRNA targeting TCGACAGCCTTATGGCGGAC in the mouse PFN1 gene Protein purification and the puromycin donor plasmid pDonor-D01 Human profilin 1 protein was purified as described (61). (Genecopoeia) for selection. Control knock-out cells were Briefly, profilin 1 plasmids were cloned between NdeI and generated using the same vectors and a scrambled EcoRI sites of pMW172, a pET derivative (62, 63) and sgRNA control targeting the sequence were expressed in Rosetta pRARE2 BL21(DE3) cells. GGCTTCGCGCCGTAGTCTTA. All constructs were Cells were grown in terrific broth to OD600 = 0.5 at 37 ºC, prepared for transfection using the GenElute HP then induced with IPTG for 3 h at 37 ºC. Pellets were Endotoxin-Free Plasmid Maxiprep Kit (Sigma-Aldrich). resuspended in 50 mM Tris HCl (pH 8.0), 10 mg/mL DNase I, 20 mg/mL PMSF, 1× protease inhibitor cocktail, RNA-seq analysis and 10 mM DTT. Cells were lysed with 150 mg/mL Total RNA was extracted from wild-type PFN1 knockout lysozyme and sonicated at 4 ºC. The cell lysate was cells (four biological replicates per condition) and total clarified by centrifugation at 20,000 × g. The supernatant

Skruber et al. | bioRχiv | Nov. 19, 2019 | 10 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. was passed over at QHighTrap column (GE Healthcare, (C56B8, 1:1000 dilution, Cell Signaling Technology); Marlborough, MA) equilibrated in 50 mM Tris-HCl (pH rabbit anti-Pan Actin (4968, 1:1000 dilution, Cell Signaling 8.0), 1 M KCl, 10 mM DTT and the flow-through Technology); rabbit anti-GAPDH (2118, 1:3000 dilution, (containing PFN1) was collected and then applied to a Cell Signaling Technology), goat anti-rabbit Alexa Fluor™ Superdex 75 (10/300) gel filtration column (GE 680 (Li-Cor) was used at 1:3500 dilution for imaging on Healthcare) equilibrated in 50 mM Tris (pH 8.0), 50 mM the Li-Cor Odyssey detection system and goat anti-rabbit KCl, 10 mM DTT. Fractions containing Profilin were HRP (Abcam) was used for X-ray detection. pooled, aliquoted, and stored at 80 ºC. Thawed Profilin aliquots were pre-cleared at 279,000 × g before use. Actin monomer/filament ratio measurements Cells were collected in lysis and F-actin stabilization DNA and protein electroporation buffer (LAS01, Cytoskeleton Inc) at 37°C in the presence The Neon Transfection System (Invitrogen) was used to of Halt Protease Inhibitor Cocktail (Thermo Fisher) and introduce DNA constructs and purified protein into cells 10mM ATP (Cytoskeleton Inc). Cells were harvested via using the 10 μL transfection kit. Briefly, cells were allowed cell scraper and incubated at 37°C for 10 minutes. to reach a confluency of 70-80%, trypsinized and pelleted Unbroken cells and debris were pelleted at room by centrifugation. The pellet was rinsed with DPBS and temperature at 250 g for 3 mins and the supernatant was resuspended in a minimum amount of buffer R then immediately centrifuged at 150,000 g at 37°C for 1 (Invitrogen) with a total of 1µg of DNA or the designated hour in a swinging bucket rotor. The supernatant was concentration of protein. Cells transfected with DNA carefully removed and the pellet was resuspended in a constructs were given 14-18 hours after transfection volume of F-actin depolymerization buffer (FAD-02, before further experimental procedures were performed. Cytoskeleton Inc.) matching the volume of the For the 0 µM concentration in protein transfections, cells supernatant. All samples were then incubated on ice for 1 were transfected with an equivalent amount of protein hour with periodic trituration and SDS buffer stained with buffer. For protein electroporation, cells were given 2.5 Orange G (40% glycerol, 6% SDS, 300 mM Tris HCl, pH hours to adhere on laminin coated coverslips before 6.8) was then added to each sample. Samples were then experiments were performed. In the combined protein analyzed by western blot. electroporation and Arp2/3 inhibition experiment, cells were given 1 hour to adhere, media was changed and Microscopy cells were treated for an additional hour with CK-666. A Most images were acquired with a Nikon A1R+ laser single 1400 v 20 ms pulse was used for both DNA and scanning confocal microscope with a GaAsP multi- protein electroporation. This protocol routinely gave > detector unit. The microscope is also equipped with total 99% transfection efficiency. internal reflection fluorescence (TIRF) and an ORCA- Flash 4.0 sCMOS camera. All confocal and TIRF imaging Western blots was performed with an Apo TIRF 60X 1.49 NA objective. Adherent cells were harvested with a cell scraper in RIPA Deconvolution-based super-resolution confocal buffer with cOmplete EDTA-free Protease Inhibitor microscopy(64) (Wilson, 2011) was performed by using Cocktail Roche (Millipore Sigma). Whole cell lysates were zoom settings higher than the Nyquist criteria, resulting in prepared by membrane disruption using repeated oversampled pixels (0.03 μm). Confocal z-stacks were passage through a 27 gauge needle. Protein content was created and then deconvolved with Nikon Elements then assessed with Pierce BCA Protein Assay Kit software using the Landweber algorithm (15 iterations, (Thermo Fisher) and diluted in SDS buffer stained with with spherical aberration correction) to create images with Orange G (40% glycerol, 6% SDS, 300 mM Tris HCl, pH approximately 150 nm resolution (65). Live cell imaging 6.8). 10µg samples were evenly loaded on an SDS-PAGE was performed using TIRF. All cells analyzed for gel (Novex 4-20% Tris-Glycine Mini Gels, Thermo Fisher, retrograde flow were non-motile. For live cell imaging, a or 15% gel as indicated). Protein was transferred to a stage incubator with CO2 and temperature control (Tokai PVDF membrane (0.2 micron, Immobilon) and blocked in Hit) was also used. Low resolution images used for 5% Bovine Serum Albumin (BSA) (Sigma-Aldrich) for 20 measuring total actin levels (Fig. 1) and electroporation mins. All antibodies were diluted in 5% BSA and 0.1% efficiency (Fig. S4) were taken with an EVOS XL digital Tween-20 (Fisher Scientific). Primary antibodies were inverted microscope objective (Life Technologies) incubated at 4°C overnight and secondary antibodies (Li- equipped with a Plan Neoflour 20X 0.5 N.A. objective. Cor; Abcam) were incubated for 2 hours at room temperature. Actin, profilin and GAPDH from whole cell Pharmacological inhibition of actin binding proteins lysate were detected with Li-Cor fluorescent antibodies on To inhibit Arp2/3, cells were treated with 50 µM of the an Odyssey detection system (Li-Cor) or via X-ray film Arp2/3 inhibitor CK-666 or its control analog CK-689 after incubation with a developing reagent (Thermo (Sigma-Aldrich) for 1 hour prior. Stock solutions of 40 mM Fisher), as indicated. WesternSure Pre-Stained CK-666 and 100mM CK-689 were prepared in DMSO. To Chemiluminescent Protein Ladder (Li-Cor) was used as a inhibit formins, cells were then treated with 10µM SMIFH2 molecular weight marker. The following (Sigma-Aldrich) or DMSO for 30 minutes. A stock solution antibodies/dilutions were used: rabbit anti-Profilin-1 of 25 mM SMIFH2 was prepared in DMSO.

Skruber et al. | bioRχiv | Nov. 19, 2019 | 11 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Immunofluorescence coverslips for 90 min. and then imaged with TIRF Cells were fixed with 4% electron microscopy grade microscopy at 1 frame/s. Images were exported into paraformaldehyde (Electron Microscopy Sciences) for ImageJ for analysis. Ten kymographs were generated per 10 min at RT and then permeabilized for 3 minutes with cell at the leading edge where fiduciary markers were 0.1% Tween-20. Cells were then washed three times with clearly visible. Retrograde flow was calculated by PBS and stained overnight at 4°C with primary antibodies measuring the distance and time that fiduciary markers diluted in PBS. They were then washed twice with PBS traveled in the kymograph and solving for rate. Two-three for 5 min, incubated with secondary antibodies (diluted biological replicates were performed per condition. 1:1000) for 1 hr at room temperature in PBS. Actin filaments were stained with Alexa Fluor 488 phalloidin or Quantification of actin and actin-binding proteins in the Alexa Fluor 568 phalloidin (diluted 1:100, Life lamellipodia: Cells labeled with Alexa568-phalloidin or Technologies) for 30 min at room temperature in immunolabeled for Arp2/3 and Mena were imaged using immunofluorescence staining buffer. Cells were washed confocal microscopy. Images were exported into ImageJ three times with PBS before mounting with Prolong for analysis. A maximum intensity projection was made Diamond (Life Technologies). The following antibodies for each confocal z-stack. Lines ten pixels in width were were used: Rabbit anti-ARPC2 (p34-Arc, EMD Millipore) drawn perpendicular to the cell edge, and fluorescence and mouse anti-Mena (clone A351F7D9, EMD Millipore) intensity was measured along the line. 20 lines were were used at a 1:500 dilution, anti-mouse IgG 647 and drawn per cell. Three biological replicates were performed anti-rabbit IgG 568 (Life Technologies) were used at per condition. 1:1000 dilution. Quantification of linear arrays in the lamellipodia: Cells Fluorescently labeling sites of active polymerization labeled with Alexa568-phalloidin were imaged using To label sites of active polymerization, cells were plated confocal deconvolution super-resolution microscopy. on coverslips that were pre-incubated with 10 µg/mL Poly- Images were deconvolved using Nikon Elements D-Lysine (Sigma-Aldrich) for one hour and then washed software and then imported into ImageJ for analysis. with PBS prior to laminin coating. Barbed ends were Confocal z-stacks were converted into a single maximum labeled using a protocol adapted from (66). Briefly, a intensity projection image and the lamellipodia of the cell stock solution of permeabilization buffer (20 mM HEPES, was manually thresholded. The “Tubeness” ImageJ 138 mM KCl, 4 mM MgCl2, 3 mM EGTA, and 1% BSA, plugin (68) was used for linear array segmentation on the pH 7.4) was prepared. Immediately prior to use, 0.025% thresholded lamellipodia. Images were convolved with a saponin and 1 mM ATP were added followed by 0.45 µM sigma value three times the minimum voxel separation. rhodamine actin (Cytoskeleton Inc). The culture medium The convolved image was binarized and the Analyze was carefully removed by pipette and enough Particle function was used to threshold objects with low permeabilization buffer was added to cover cells. After 1 circularity (0.0-0.3) and an area measurement greater minute, permeabilization buffer was gently removed by than 0.1 µm. Filament density was calculated by pipette, rinsed briefly with 1X PBS and immediately fixed normalizing the number of segmented arrays to the in 4% paraformaldehyde in cytoskeleton stabilization perimeter of the cell. Three biological replicates were buffer (67) for 10 min. After carefully washing in 1X PBS performed per condition. 3 times, samples were incubated with phalloidin-488 or phalloidin-568 (diluted 1:100, Life Technologies) for 20 Statistical Analysis min at room temperature. Cells were imaged using Statistical significance was assessed by either a two- identical conditions for comparison. tailed student’s t-test (comparison of two conditions) or by ANOVA followed by Tukey’s post hoc test (comparison of Image Analysis three or more conditions) using Graphpad Prism software. Quantification of total actin filaments per cell: Cells were Dunn’s post hoc test was used in place of Tukey’s with transfected with GFP or a GFP-PFN1 construct, and then nonparametric datasets, as indicated. fixed and labeled with Alexa568-phalloidin. Images were taken on the EVOS XL microscope using identical Conflict of Interest illumination and camera exposure conditions. Image files The authors declare no conflict of interest. were imported into ImageJ, the background was subtracted, and cells were segmented by fluorescence Acknowledgements intensity-based thresholding using the phalloidin channel. We would like to thank Dr. Alpha Yap and Suzie Verma Mean GFP and Alexa568-phalloidin values were used to for the generous gift of the EGFP-FP4-Mito and EGFP- assess relative PFN1 and F-actin levels in each cell. At AP4-Mito DNA constructs. This project was supported by least three biological replicates were performed per a National Institutes of Health (NIH) Maximizing condition. Investigators' Research Award for Early Stage Investigators (R35GM133485) to J.L.H.-R. and an NIH Quantification of retrograde flow: Cells transfected with Pathway to Independence Award (R00 NS087104) to Lifeact-mRuby were plated onto laminin-coated E.A.V. Skruber et al. | bioRχiv | Nov. 19, 2019 | 12 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

References 16. P. C. Tseng, M. S. Runge, J. A. Cooper, R. C. Williams, T. D. Pollard, Physical, immunochemical, and 1. D. J. Gordon, J. L. Boyer, E. D. Korn, functional properties of Acanthamoeba profilin. J. Cell Comparative biochemistry of non-muscle . J. Biol. Biol. 98, 214–221 (1984). Chem. 252, 8300–8309 (1977). 17. A. J. Bestul et al., Fission yeast profilin is tailored 2. S. A. Koestler et al., F- and G-actin to facilitate actin assembly by the cytokinesis formin concentrations in lamellipodia of moving cells. PLoS One. Cdc12. Mol. Biol. Cell. 26, 283–293 (2015). 4, e4810 (2009). 18. D. R. Kovar, J. R. Kuhn, A. L. Tichy, T. D. Pollard, 3. L. G. Tilney, The polymerization of actin. II. How The fission yeast cytokinesis formin Cdc12p is a barbed nonfilamentous actin becomes nonrandomly distributed in end actin filament capping protein gated by profilin. J. Cell sperm: evidence for the association of this actin with Biol. 161, 875–887 (2003). membranes. J. Cell Biol. 69, 51–72 (1976). 19. D. R. Kovar, E. S. Harris, R. Mahaffy, H. N. Higgs, 4. V. K. Vinson, E. M. De La Cruz, H. N. Higgs, T. T. D. Pollard, Control of the assembly of ATP- and ADP- D. Pollard, Interactions of Acanthamoeba profilin with actin by formins and profilin. Cell. 124, 423–435 (2006). actin and nucleotides bound to actin. Biochemistry. 37, 10871–10880 (1998). 20. S. Romero et al., Formin is a processive motor that requires profilin to accelerate actin assembly and 5. K. Skruber, T.-A. Read, E. A. Vitriol, associated ATP hydrolysis. Cell. 119, 419–429 (2004). Reconsidering an active role for G-actin in cytoskeletal regulation. J. Cell Sci. 131 (2018), 21. D. Vavylonis, D. R. Kovar, B. O’Shaughnessy, T. doi:10.1242/jcs.203760. D. Pollard, Model of formin-associated actin filament elongation. Mol. Cell. 21, 455–466 (2006). 6. C. Suarez et al., Profilin regulates F-actin network homeostasis by favoring formin over Arp2/3 complex. 22. S. D. Hansen, R. D. Mullins, VASP is a Dev. Cell. 32, 43–53 (2015). processive actin polymerase that requires monomeric actin for barbed end association. J. Cell Biol. 191, 571– 7. J. D. Rotty et al., Profilin-1 serves as a 584 (2010). gatekeeper for actin assembly by Arp2/3-dependent and -independent pathways. Dev. Cell. 32, 54–67 (2015). 23. L. Pasic, T. Kotova, D. A. Schafer, Ena/VASP proteins capture actin filament barbed ends. J. Biol. 8. C. W. Lee et al., Dynamic localization of G-actin Chem. 283, 9814–9819 (2008). during membrane protrusion in neuronal motility. Curr. Biol. 23, 1046–1056 (2013). 24. L. M. Machesky et al., Scar, a WASp-related protein, activates nucleation of actin filaments by the 9. W. Lei et al., Phosphoinositide-dependent Arp2/3 complex. Proc. Natl. Acad. Sci. USA. 96, 3739– enrichment of actin monomers in dendritic spines 3744 (1999). regulates synapse development and plasticity. J. Cell Biol. 216, 2551–2564 (2017). 25. L. Blanchoin, T. D. Pollard, R. D. Mullins, Interactions of ADF/cofilin, Arp2/3 complex, capping 10. E. A. Vitriol et al., Two functionally distinct protein and profilin in remodeling of branched actin sources of actin monomers supply the leading edge of filament networks. Curr. Biol. 10, 1273–1282 (2000). lamellipodia. Cell Rep. 11, 433–445 (2015). 26. D. Yarar, J. A. D’Alessio, R. L. Jeng, M. D. Welch, 11. B. J. Perrin, J. M. Ervasti, The actin gene family: Motility determinants in WASP family proteins. Mol. Biol. function follows isoform. Cytoskeleton (Hoboken). 67, Cell. 13, 4045–4059 (2002). 630–634 (2010). 27. D. Chereau et al., Actin-bound structures of 12. M. Kapustina, T.-A. Read, E. A. Vitriol, Wiskott-Aldrich syndrome protein (WASP)-homology Simultaneous quantification of actin monomer and domain 2 and the implications for filament assembly. filament dynamics with modeling-assisted analysis of Proc. Natl. Acad. Sci. USA. 102, 16644–16649 (2005). photoactivation. J. Cell Sci. 129, 4633–4643 (2016). 28. P. Bieling et al., WH2 and proline-rich domains of 13. J. R. Terman, A. Kashina, Post-translational WASP-family proteins collaborate to accelerate actin modification and regulation of actin. Curr. Opin. Cell Biol. filament elongation. EMBO J. 37, 102–121 (2018). 25, 30–38 (2013). 29. M. Barzik et al., Ena/VASP proteins enhance 14. T. A. Burke et al., Homeostatic actin cytoskeleton actin polymerization in the presence of barbed end networks are regulated by assembly factor competition for capping proteins. J. Biol. Chem. 280, 28653–28662 monomers. Curr. Biol. 24, 579–585 (2014). (2005). 15. D. A. Kaiser, V. K. Vinson, D. B. Murphy, T. D. 30. D. Chereau, R. Dominguez, Understanding the Pollard, Profilin is predominantly associated with role of the G-actin-binding domain of Ena/VASP in actin monomeric actin in Acanthamoeba. J. Cell Sci. 112 ( Pt assembly. J. Struct. Biol. 155, 195–201 (2006). 21), 3779–3790 (1999).

Skruber et al. | bioRχiv | Nov. 19, 2019 | 13 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

31. J. B. Moseley, B. L. Goode, The yeast actin 45. D. Breitsprecher et al., Clustering of VASP cytoskeleton: from cellular function to biochemical actively drives processive, WH2 domain-mediated actin mechanism. Microbiol. Mol. Biol. Rev. 70, 605–645 filament elongation. EMBO J. 27, 2943–2954 (2008). (2006). 46. L. Mohapatra, T. J. Lagny, D. Harbage, P. R. 32. T. M. Svitkina, G. G. Borisy, Arp2/3 complex and Jelenkovic, J. Kondev, The Limiting-Pool Mechanism actin depolymerizing factor/cofilin in dendritic Fails to Control the Size of Multiple Organelles. Cell Syst. organization and of actin filament array in 4, 559–567.e14 (2017). lamellipodia. J. Cell Biol. 145, 1009–1026 (1999). 47. J. E. Bear et al., Negative regulation of fibroblast 33. J. E. Bear et al., Antagonism between Ena/VASP motility by Ena/VASP proteins. Cell. 101, 717–728 (2000). proteins and actin filament capping regulates fibroblast 48. B. J. Nolen et al., Characterization of two classes motility. Cell. 109, 509–521 (2002). of small molecule inhibitors of Arp2/3 complex. Nature. 34. C. Yang et al., Novel roles of formin mDia2 in 460, 1031–1034 (2009). lamellipodia and filopodia formation in motile cells. PLoS 49. S. M. Janicki et al., From silencing to gene Biol. 5, e317 (2007). expression. Cell. 116, 683–698 (2004). 35. S. A. Koestler, S. Auinger, M. Vinzenz, K. 50. Y. Ben-Ari et al., The life of an mRNA in space Rottner, J. V. Small, Differentially oriented populations of and time. J. Cell Sci. 123, 1761–1774 (2010). actin filaments generated in lamellipodia collaborate in pushing and pausing at the cell front. Nat. Cell Biol. 10, 51. T. M. Svitkina et al., Mechanism of filopodia 306–313 (2008). initiation by reorganization of a dendritic network. J. Cell Biol. 160, 409–421 (2003). 36. E. A. Vitriol, A. L. Wise, M. E. Berginski, J. R. Bamburg, J. Q. Zheng, Instantaneous inactivation of 52. M. Lorenz, H. Yamaguchi, Y. Wang, R. H. Singer, cofilin reveals its function of F-actin disassembly in J. Condeelis, Imaging sites of N-wasp activity in lamellipodia. Mol. Biol. Cell. 24, 2238–2247 (2013). lamellipodia and invadopodia of carcinoma cells. Curr. Biol. 14, 697–703 (2004). 37. W. Witke, J. D. Sutherland, A. Sharpe, M. Arai, D. J. Kwiatkowski, Profilin I is essential for cell survival and 53. F. Korobova, T. Svitkina, Arp2/3 complex is cell division in early mouse development. Proc. Natl. important for filopodia formation, growth cone motility, and Acad. Sci. USA. 98, 3832–3836 (2001). neuritogenesis in neuronal cells. Mol. Biol. Cell. 19, 1561– 1574 (2008). 38. O. C. Ezezika et al., Incompatibility with formin Cdc12p prevents human profilin from substituting for 54. A. K. Mongiu, E. L. Weitzke, O. Y. Chaga, G. G. fission yeast profilin: insights from crystal structures of Borisy, Kinetic-structural analysis of neuronal growth fission yeast profilin. J. Biol. Chem. 284, 2088–2097 cone veil motility. J. Cell Sci. 120, 1113–1125 (2007). (2009). 55. Y. Fan et al., Stimulus-dependent 39. J. Pernier, S. Shekhar, A. Jegou, B. Guichard, M.- phosphorylation of profilin-1 in angiogenesis. Nat. Cell F. Carlier, Profilin Interaction with Actin Filament Barbed Biol. 14, 1046–1056 (2012). End Controls Dynamic Instability, Capping, Branching, 56. J. D. Thomas et al., Disrupted prenatal RNA and Motility. Dev. Cell. 36, 201–214 (2016). processing and myogenesis in congenital myotonic 40. P. J. Goldschmidt-Clermont, L. M. Machesky, S. dystrophy. Genes Dev. 31, 1122–1133 (2017). K. Doberstein, T. D. Pollard, Mechanism of the interaction 57. J. Wu, O. Anczuków, A. R. Krainer, M. Q. Zhang, of human platelet profilin with actin. J. Cell Biol. 113, C. Zhang, OLego: fast and sensitive mapping of spliced 1081–1089 (1991). mRNA-Seq reads using small seeds. Nucleic Acids Res. 41. D. Safer, M. Elzinga, V. T. Nachmias, Thymosin 41, 5149–5163 (2013). beta 4 and Fx, an actin-sequestering peptide, are 58. Y. Zhang et al., An RNA-sequencing indistinguishable. J. Biol. Chem. 266, 4029–4032 (1991). transcriptome and splicing database of glia, neurons, and 42. D. Van Goor, C. Hyland, A. W. Schaefer, P. vascular cells of the cerebral cortex. J. Neurosci. 34, Forscher, The role of actin turnover in retrograde actin 11929–11947 (2014). network flow in neuronal growth cones. PLoS One. 7, 59. M. D. Robinson, D. J. McCarthy, G. K. Smyth, e30959 (2012). edgeR: a Bioconductor package for differential 43. M. H. Symons, T. J. Mitchison, Control of actin expression analysis of digital gene expression data. polymerization in live and permeabilized fibroblasts. J. Bioinformatics. 26, 139–140 (2010). Cell Biol. 114, 503–513 (1991). 60. Y. Qi, J. K. Wang, M. McMillian, D. M. 44. M. Edwards et al., Capping protein regulators Chikaraishi, Characterization of a CNS cell line, CAD, in fine-tune actin assembly dynamics. Nat. Rev. Mol. Cell which morphological differentiation is initiated by serum Biol. 15, 677–689 (2014). deprivation. J. Neurosci. 17, 1217–1225 (1997).

Skruber et al. | bioRχiv | Nov. 19, 2019 | 14 bioRxiv preprint doi: https://doi.org/10.1101/849356; this version posted November 20, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

61. J. L. Henty-Ridilla, M. A. Juanes, B. L. Goode, Profilin Directly Promotes Growth through Residues Mutated in Amyotrophic Lateral Sclerosis. Curr. Biol. 27, 3535–3543.e4 (2017). 62. S. Boopathy et al., Structural basis for mutation- induced destabilization of profilin 1 in ALS. Proc. Natl. Acad. Sci. USA. 112, 7984–7989 (2015). 63. J. C. Eads et al., Structure determination and characterization of Saccharomyces cerevisiae profilin. Biochemistry. 37, 11171–11181 (1998). 64. T. Wilson, Resolution and optical sectioning in the confocal microscope. J. Microsc. 244, 113–121 (2011). 65. Z. Osking et al., ALS-Linked SOD1 Mutants Enhance Neurite Outgrowth and Branching in Adult Motor Neurons. iScience. 11, 294–304 (2019). 66. A. Y. Chan et al., EGF stimulates an increase in actin nucleation and filament number at the leading edge of the lamellipod in mammary adenocarcinoma cells. J. Cell Sci. 111 ( Pt 2), 199–211 (1998). 67. J. V. Small, J. E. Celis, Filament arrangements in negatively stained cultured cells: the organization of actin. Cytobiologie. 16, 308–325 (1978). 68. Y. Sato et al., Three-dimensional multi-scale line filter for segmentation and visualization of curvilinear structures in medical images. Med Image Anal. 2, 143– 168 (1998).

Skruber et al. | bioRχiv | Nov. 19, 2019 | 15